A. UV–Vis absorption

Figure 1 shows the UV–Vis absorption spectra for each perovskite film, where it can be seen that each sample exhibits a relatively strong absorption peak centered around 500 nm. For our control sample CH₃NH₃PbI_3−x Cl _x , its optical absorption extend to long wave length >750 nm where there is a distinct edge which when extrapolated meets the x-axis at approximately 820 nm, equating to a band gap of 1.51 eV in reasonable agreement with earlier reports. ^{Reference Lee, Teuscher, Miyasaka, Murakami and Snaith6,Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca21} This is close to the value reported for CH₃NH₃PbI₃, ^{Reference Kulkarni, Baikie, Boix, Yantara, Mathews and Mhaisalkar22,Reference Bi, Shao, Yuan, Xiao, Wang, Gao and Huang23} again suggesting the low/trace levels of Cl in this perovskite. Among the mixed metal perovskites only CH₃NH₃(Pb_0.75Bi_0.25)I_3−x Cl _x showed a weaker absorption shoulder in the near-IR region of the light spectrum, at a wave length similar to that observed for the control sample (>750 nm). We tentatively attribute this feature to the presence of pure lead perovskite phases within the sample, given the 3× difference in lead molar content compared to bismuth in the precursor. Further increases in bismuth content within the sample, from Pb_0.75Bi_0.25 to Pb_0.25Bi_0.75, results in the eventual disappearance of this feature. For the pure bismuth perovskites, CH₃NH₃BiI₃ (1:1) prepared from a 1:1 precursor stoichiometry shows a weak absorption edge at approximately 645 nm, whereas CH₃NH₃BiI₃ (3:1) prepared from a 3:1 precursor stoichiometry and CH₃NH₃BiI_3−x Cl _x do not. For CH₃NH₃BiI_3−x Cl _x we estimate an optical band gap of 2.1 eV via extrapolation of the absorption edge beyond 510 nm. Note however that this method does not readily extend to every perovskite considered here, as the slow and incomplete drop-off in absorption (likely an indication of pronounced light scatter) will introduce unintended artifacts. ^{Reference Tian and Scheblykin24}

FIG. 1.

Absorbance spectra of CH₃NH₃PbI_3−x Cl _x , CH₃NH₃BiI_3−x Cl _x , CH₃NH₃BiI₃ (1:1), CH₃NH₃BiI₃ (3:1), CH₃NH₃(Pb_0.25Bi_0.75)I_3−x Cl _x , CH₃NH₃(Pb_0.5Bi_0.5)I_3−x Cl _x and CH₃NH₃(Pb_0.75Bi_0.25)I_3−x Cl _x films prepared on glass substrates. Precursor samples were annealed at 90 °C for 90 min to form each perovskite.

B. SEM/EDS characterization

To study in detail the microstructure of each perovskite film, in addition to confirming their composition, we recorded a series of SEM images and EDS chemical surveys. Figure 2 shows representative SEM images for each perovskite. From Fig. 2(a), which corresponds to our control sample, it can be seen that CH₃NH₃PbI_3−x Cl _x consists of a relatively dense distribution of elongated crystals, each having a characteristic length between 3 and 10 µm. Quantitative analysis of this image suggests a surface coverage of 74%. For the mixed metal perovskites, shown here in Figs. 2(b)–2(d), two distinct structure types are present; a compact distribution of micron-sized crystallites that cover a sizable fraction of the substrate, and dispersed structures with a size between 20 and 30 µm. Note that these relatively large features are often surrounded by a region void of any material, suggesting that mass transport and/or substrate de-wetting has taken place during their formation. ^{Reference Ondarçuhu and Aimé25} In general, these structures are present in greater number when the relative concentration of Bi in the sample increases. Although this might suggest the presence of separate perovskite phases in the sample (i.e., regions of pure CH₃NH₃PbI_3−x Cl _x that are spatially distinct from regions of pure CH₃NH₃BiI_3−x Cl _x ), EDS mapping indicates that the distribution of Bi throughout the sample is reasonably uniform (data presented in the Supplementary Material). As a consequence we suggest that the formation of either crystallite structure is not determined primarily by metal content, and is instead governed by factors that include halide composition, the dynamics of precursor conversion and the local presence of defects/impurities in the sample that act as possible nucleation sites for crystal formation.

FIG. 2.

SEM images of (a) CH₃NH₃PbI_3−x Cl _x , (b) CH₃NH₃(Pb_0.75Bi_0.25)I_3−x Cl _x , (c) CH₃NH₃(Pb_0.5Bi_0.5)I_3−x Cl _x , (d) CH₃NH₃(Pb_0.25Bi_0.75)I_3−x Cl _x , (e) CH₃NH₃BiI_3−x Cl _x , (f) CH₃NH₃BiI₃ (1:1), and (g) CH₃NH₃BiI₃ (3:1) films prepared on silicon substrates. Scale bar corresponds to 10 µm.

The SEM image for CH₃NH₃BiI_3−x Cl _x is shown in Fig. 2(e). Here it can be seen that multiple ‘flower’ shaped features are present, many of which appear to sit edge on from the substrate surface. These features are often surrounded by regions of what appears to be less structured material, however their total surface coverage is low (c. 50%). In Fig. 3(f) the film corresponding to CH₃NH₃BiI₃ (1:1) shows a mixture of short needle like crystals protruding from central points (about 10 µm long) sitting on top of round unstructured regions that are between 10 and 35 µm in diameter. For this sample the total surface coverage is much higher than for CH₃NH₃BiI_3−x Cl _x . The CH₃NH₃BiI₃ (3:1) perovskite film is principally characterized by large crystals (>50 µm in length) alongside local regions containing smaller structures as shown in Fig. 2(g). Connectivity between these regions is however poor, with many areas of the substrate containing no material whatsoever.

FIG. 3.

SEM image of (a) CH₃NH₃BiI_3−x Cl _x and corresponding EDS analysis with elemental mapping of (b) carbon, (c) nitrogen, (d) bismuth, (e) iodine, and (f) chlorine.

Having established that the bismuth perovskite films are comprised of multiple crystal structures with characteristically distinct sizes, we proceed to discuss their chemical composition as determined by EDS mapping. Table I summarizes the results of these measurements, where values correspond to the stoichiometric ratio of each element. Corresponding images are presented in the Supplementary Material Figs. S1–S7. As reported previously, Cl was almost completely absent in the CH₃NH₃PbI_3−x Cl _x film, ^{Reference Zhao and Zhu26} indicating its removal from the sample during processing (i.e., via sublimation with excess MAI) ^{Reference Yu, Wang, Xie, Li, Chen and Zhao27–Reference Zhang, Zhang, Yin, Jiang, Liu, Meng, Zhao and Wang29} or its preferential location at the substrate interface. As the ratio of Bi is increased in the mixed metal perovskite the Cl content rises from 0.46 (25% Bi) up to 1.47 (100% Bi). Spatially averaged EDS data for CH₃NH₃BiI_3−x Cl _x indicates a halide ratio of approximately 2.5:1.5I:Cl, however this result masks important local variations that will be discussed later. In line with the choice of starting precursor stoichiometries, CH₃NH₃BiI₃ (3:1) shows the highest concentration of iodine. Although carbon was found to exist in a 1:1 ratio with nitrogen during elemental analysis of MAI, its relative concentration in each film is increased. We attribute this enhanced carbon content to a combination of sample contamination by carbon caused by electron beam exposure in the SEM and residual levels of trapped solvent.

TABLE I.

The stoichiometric ratio of elements as determined by EDS for the CH₃NH₃PbI_3−x Cl _x , mixed metal perovskites, CH₃NH₃BiI _x Cl _y , and CH₃NH₃BiI _x films as determined by EDS. The numbers in parenthesize correspond to the target stoichiometric ratio determined by the composition of the precursor solutions used.

To identify the origin(s) for the different crystal structures present in the mixed Pb/Bi perovskite films, we note that EDS maps show a generally homogenous distribution of these specific elements throughout each sample. Cl and I are also similarly distributed in both the large and small structures. In contrast, from Fig. 3 noticeable differences in the distribution of the halides in the CH₃NH₃BiI_3−x Cl _x film are determined, where I closely maps to the large flower shaped structures present and Cl maps to the thin layer than lies beneath. It appears therefore that in these films there are two different stoichiometries which we subsequently refer to as I-rich and Cl-rich. To determine their molar ratio in each region we conducted a series of point EDS chemical surveys.

As summarized in Table II, in the I-rich regions bismuth preferentially binds to iodine when forming crystalline structures. Unlike in the lead-halide perovskite, a significant amount of chlorine is present in the film, suggesting that its removal during perovskite formation is less efficient when bismuth is present. In the I-rich regions the Bi:Cl:I ratio 1:0.67:2.67. Within the Cl-rich regions the Bi:Cl:I ratio is 1:1.37:1.64. These ratios indicate that the bismuth to halide ratio in both cases is almost 1:3, as one would expect for an ABX₃ perovskite.

TABLE II.

The EDS point analysis for the CH₃NH₃BiI _x Cl _y film highlighting the distribution of iodine and chlorine rich regions areas.

C. GIWAXS characterization

Given the difference in ionic radii for lead, bismuth, chlorine, and iodine, we suggest that the various structures imaged using SEM will also have different crystal lattice parameters. To test this hypothesis, and to also verify the presence of crystallites with a perovskite unit cell structure, samples were measured using GIWAXS. Radially integrated 1D X-ray scattering patterns from each sample are shown in Fig. 4, alongside patterns from films prepared using the precursor materials MAI, BiCl₃, and BiI₃. Immediately it can be seen that the diffraction pattern of each perovskite is not a simple linear combination of its precursor ingredients, thereby confirming the precursor components have combined to form new crystalline phases. The GIWAXS data for CH₃NH₃BiI_3−x Cl _x shown in Fig. 4(a) demonstrates minimal evidence for X-ray scattering from MAI and BiCl₃, supporting the idea that the perovskite has a stoichiometry close to its target value. Nevertheless the presence of X-ray scattering peaks other than those attributable to the perovskite suggests more than one crystalline phase has been formed. For CH₃NH₃BiI₃ (1:1) and CH₃NH₃BiI₃ (3:1), there are clear indications that unreacted BiI₃ is present within the film as well as unreacted MAI. This result implies that for these blends the precursors do not undergo complete conversion possibly because they are not stoichiometrically well matched. Despite the different mixing ratios used the X-ray scatter pattern from both these new crystal phases are very similar. This is in stark contrast to the different structures observed using SEM (see Fig. 3), demonstrating a strong sensitivity of sample microstructure to precursor composition. This sensitivity to composition has been noted by others. ^{Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca21,Reference Yu, Wang, Xie, Li, Chen and Zhao27}

FIG. 4.

(a) The 1D radially integrated GIWAXS patterns of MAI, BiCl₃, BiI₃, CH₃NH₃BiI_3−x Cl _x , CH₃NH₃BiI₃ (1:1), and CH₃NH₃BiI₃ (3:1) on the silicon substrates and (b) The 1D radially integrated GIWAXS patterns of CH₃NH₃PbI_3−x Cl _x , CH₃NH₃(Pb_0.75Bi_0.25)I_3−x Cl _x , CH₃NH₃(Pb_0.5Bi_0.5)I_3−x Cl _x , CH₃NH₃(Pb_0.25Bi_0.75)I_3−x Cl _x , CH₃NH₃BiI_3−x Cl _x , CH₃NH₃BiI₃ (1:1), and CH₃NH₃BiI₃ (3:1) films prepared on silicon substrates.

The X-ray scattering from CH₃NH₃PbI_3−x Cl _x is similar to previously published data, ^{Reference Colella, Mosconi, Fedeli, Listorti, Gazza, Orlandi, Ferro, Besagni, Rizzo, Calestani, Gigli, De Angelis and Mosca21,Reference Tan, Moore, Saliba, Sai, Estroff, Hanrath, Snaith and Wiesner30,Reference Park, Philippe, Gustafsson, Sveinbjornsson, Hagfeldt, Johansson and Boschloo31} allowing us attribute the scattering peaks at q = 0.99 to 1.00 Å⁻¹, q = 1.41 to 1.42 Å⁻¹, and q = 1.99 to 2.00 Å⁻¹ to X-ray scatter from the (002) and (110), (112) and (200), and (004) and (220) features of tetragonal CH₃NH₃PbI_3−x Cl _x respectively. For this system the corresponding lattice parameters are a = b = 8.85 Å and c = 12.69 Å, close to the values reported for the pure tri-iodide perovskite. ^{Reference Baikie, Fang, Kadro, Schreyer, Wei, Mhaisalkar, Graetzel and White32–Reference Stoumpos, Malliakas and Kanatzidis34} The GIWAXS pattern for CH₃NH₃(Pb_0.75Bi_0.25)I_3−x Cl _x is broadly similar to CH₃NH₃PbI_3−x Cl _x , albeit with the addition of new scattering peaks at q = 0.89 Å⁻¹, q = 1.21 Å⁻¹ and peaks of relatively weak intensity between 1.50 Å⁻¹ and 2.00 Å⁻¹. This is consistent with the UV–Vis data presented in Fig. 1 that indicated a similar absorption onset for the two samples. For CH₃NH₃(Pb_0.5Bi_0.5)I_3−x Cl _x , the relative intensity of the peaks at 0.89 Å⁻¹ and 1.21 Å⁻¹ is higher, however the peaks corresponding to the (110) and (220) reflections from CH₃NH₃PbI_3−x Cl _x are found to undergo both a reduction in relative intensity alongside a small shift (c. 1%) to higher q values. This not only implies a reduction in Pb-rich perovskite content within the sample, consistent with the sample preparation conditions and UV–Vis/EDS data discussed earlier, but also a contraction in the unit cell size that may result from compressive strain. Further reductions in Pb content within the sample (i.e., CH₃NH₃(Pb_0.25Bi_0.75)I_3−x Cl _x ), results in an X-ray scattering pattern that is consistent with the observed trend.

In agreement with the SEM data presented in Fig. 2, the presence of X-ray scattering peaks that can be partly attributed to CH₃NH₃PbI_3−x Cl _x in the mixed metal samples suggests the formation of distinct crystal phases rather than the formation of an alloy. We suggest that prior to mixing precursor solutions for CH₃NH₃PbI_3−x Cl _x and CH₃NH₃BiI_3−x Cl _x , the partial formation of metal halide complexes has already taken place that favors the eventual formation of crystals containing either (but not both) metals. Interestingly, the observed peaks in the CH₃NH₃PbI_3−x Cl _x GIWAXS pattern at 0.89 Å⁻¹ and 1.90 Å⁻¹ match those from BiI₃, despite the fact that BiI₃ was not present as precursor material. It appears therefore that for these mixed Pb/Bi samples significant halide ion exchange has taken place during sample processing that results in the formation of BiI₃. Furthermore, from Fig. 4(b) it can also be seen that the mixed Pb/Bi samples follow a trend that more closely resembles a transition from CH₃NH₃PbI₃ to CH₃NH₃BiI₃, rather than to CH₃NH₃BiI_3−x Cl _x , even though the precursors for CH₃NH₃BiI_3−x Cl _x were used for mixing, further indicating that halide exchange occurs. ^{Reference Yoon, Stamplecoskie and Kamat35}

To better understand the origin of each new feature observed in our GIWAXS measurements, X-ray scattering patterns of single phase powder samples were simulated the using CrystalDiffract 6 software package. ³⁶ Crystal structures (.cif files) for CH₃NH₃PbI₃ and BiI₃ were taken from a previous report. ^{Reference Stoumpos, Malliakas and Kanatzidis34} The CH₃NH₃PbI₃ scattering pattern measured was in good agreement with the previously reported structure. Crystal structures for bismuth perovskites have not been widely reported; therefore we started from a generic ABX₃ structure file (A = methylammonium, B = Bi, and X = halide) and iteratively adjusted the lattice parameters until it matched the experimental data. The fitted lattice parameters for CH₃NH₃BiI₃ (prepared from both the 1:1 and 3:1 precursor mixtures) correspond to a tetragonal structure. Figure 5 shows the simulated GIWAXS pattern of a perovskite with unit cell dimensions a = b = 8.67 Å and c = 12.41 Å and a mixture of this simulated pattern BiI₃ and MAI. This mixture is compared to the measured data for CH₃NH₃BiI₃ and as shown, there is reasonable agreement between the mixture and the experimentally measured data. Specifically, the simulated ABX₃ pattern shows two peaks at the q = 1.01 Å⁻¹ and 1.02 Å⁻¹, corresponding to the (002) and (110) reflections respectively. The experimental pattern for CH₃NH₃BiI₃ also shows a peak at the q = 1.02 Å⁻¹, alongside a shoulder at q = 1.01 Å⁻¹.

FIG. 5.

The GIWAXS patterns from (a) the simulated ABX₃ (tetragonal a = b = 8.67 Å and c = 12.41 Å), simulated mixture of ABX₃, MAI, and BiI₃, and the experimentally observed pattern from CH₃NH₃BiI₃ (3:1) and (b) the simulated ABX₃ (tetragonal a = b = 8.67 Å and c = 12.41 Å), simulated mixture of ABX₃, MAI, BiCl₃, and BiI₃, and the experimentally observed pattern from CH₃NH₃BiI_3−x Cl _x .

The mixture and the experimental pattern both show several peaks between q = 1.5–2.3 Å⁻¹. Therefore, the relatively intense peaks observed can be modeled using a mixture of a generic ABX₃ crystal lattice, MAI and BiI₃, allowing us to conclude that CH₃NH₃BiI₃ forms a perovskite structure. However, between q = 0.5 Å⁻¹ and 1 Å⁻¹ there are several weak features that cannot be readily identified, alongside a peak showed at q = 1.21 Å⁻¹ that is present in the GIWAXS data for every bismuth containing perovskite. These features are tentatively attributed to the presence of unreacted BiCl₃ and other impurities.

We also adapted our model to simulate the X-ray scattering pattern for CH₃NH₃BiI_3−x Cl _x , based on a generic ABX₃ structure, and combined it with data for MAI, BiI₃, and BiCl₃. The results are presented in Fig. 5(b). To match the experimental data the best fit lattice parameters used in the simulated ABX₃ structure were a = b = 8.29 Å and c = 11.89 Å. Weak X-ray scattering features at q = 1.21, 1.64, and 2.04 Å⁻¹ were not well matched by our model suggesting the presence of unidentified residual phases that are in addition to the perovskite, MAI, BiI₃ and BiCl₃.

D. Solar cell performance

Our measurements on methylammonium bismuth- and methylammonium mixed Pb/Bi-perovskite films have evidenced systems with limited phase purity (relative to our control sample for CH₃NH₃PbI_3−x Cl _x ), in addition to the presence of crystalline structures that have incomplete surface coverage and a range of length-scales. To determine whether such films exhibit any photovoltaic functionality, we fabricated a series of inverted architecture solar cell devices according to the structure ITO/PEDOT:PSS/Perovskite/PCBM/Ca/Al. Light J–V curves and tabulated metrics for each device are shown in Fig. 6 and Table III respectively. Here it can be seen that as expected the control devices based on CH₃NH₃PbI_3−x Cl _x exhibited the best solar cell performance by a considerable margin, recording average efficiencies of 8.25% (J _SC = −14.7 mA/cm², V _OC = 0.86 V, FF = 65.3%). This result is similar to our previous work using the same materials and processing routes. ^{Reference Barrows, Pearson, Kwak, Dunbar, Buckley and Lidzey37} In comparison, and in agreement with Park et al. ^{Reference Park, Philippe, Zhang, Rensmo, Boschloo and Johansson19} PCE values for the Bi containing devices were very low. Mixed Pb/Bi perovskite solar cells exhibited J _SC and V _OC below 0.03 mA/cm² and 0.5 V respectively. For the CH₃NH₃BrI_3−x Cl _x solar cells [data shown in Fig. 6(b)] a maximum PCE of 0.009% was determined (J _SC of −0.03 mA/cm², V _OC = 0.78 V, FF = 44.8%). Although its performance was poor, the J–V curve shows the distinctive shape for a photovoltaic device; furthermore a V _OC of 0.78 V was high compared to the other Bi containing perovskite solar cells. Lastly, for CH₃NH₃BrI₃, a relatively high J _SC of 0.08 mA/cm² was measured when the perovskite was prepared from a 1:1 mixture of precursor materials. This was offset however by a low V _OC value of 0.22 V. CH₃NH₃BrI₃ prepared from a 3:1 precursor mixture did not show any measureable photovoltaic effect.

FIG. 6.

J–V curves for solar cell devices based on CH₃NH₃PbI_3−x Cl _x , CH₃NH₃(Pb_0.25Bi_0.75)I_3−x Cl _x , CH₃NH₃(Pb_0.5Bi_0.5)I_3−x Cl _x , CH₃NH₃(Pb_0.75Bi_0.25)I_3−x Cl _x , and CH₃NH₃BiI_3−x Cl _x .

TABLE III.

Photovoltaic performance parameters of the devices produced for this study.

Despite the disappointing device efficiencies reported for the Bi perovskite solar cells it is still noteworthy that these materials can produce a measurable photovoltaic effect. From our SEM, EDS, and GIWAXS data some of the reasons why the performance of these devices are poor can be understood. The material characterization data suggests that both the mixed Pb/Bi and the 100% bismuth based perovskite films contain a significant amount of impurity phases from compounds including BiI₃ and MAI. It is assumed that these are responsible for the dramatically reduced J _SC, V _OC, and FF values relative to our control device as they may act to increase the internal resistance of the devices. In addition, the inhomogeneous surface coverage of these perovskites will also limit photocurrent efficiency due a reduced effective active area. Given the promising V _OC value for CH₃NH₃BrI_3−x Cl _x , we believe sufficient incentive exists to continue our studies into this material with a focus on the optimization of the semiconductor phase purity and microstructure. By eliminating impurities from the material as well as maximizing its coverage on a substrate layer (potentially via a refined precursor conversion process), it should be possible to realize hybrid perovskite solar cells with improved quantum efficiencies and PCEs, thereby providing a potential low-toxicity alternative to lead-based perovskites for solar cell applications.